🎯 Information is never distributed evenly, and a lot of money is made in exactly that gap. Reading one deal takes three questions: What does each side know? Who can verify more easily? Who can prove themselves at a cost a counterfeit cannot pay?
The seller knows the product better than the buyer. The boss knows the company’s cash flow better than the employee, and the employee knows what the work actually feels like better than the boss does. The platform knows the traffic algorithm better than the merchants. The doctor knows medicine better than the patient. Insiders know the company better than retail investors do.
This situation, where “one side holds information the other cannot see,” is called information asymmetry in economics. It is not a morality problem. It is the default state of almost every transaction, and it is the shared foundation underneath modern microeconomics, business model design and game theory.
1. Information asymmetry comes in two kinds, and they happen at different times
1. Hidden attributes before the deal: adverse selection
The other side has already hidden “what I really am” before the deal is struck. Hidden damage in a used car, a job applicant’s real ability, a policyholder’s known medical history — all of these belong here. Economics calls it adverse selection.
2. Hidden actions after the deal: moral hazard
The contract is signed, the money is paid, and then the other side starts doing things you cannot see. The rental car that is overloaded every day, the funded startup that spends the budget somewhere else, the driver who fully insured the car and now drives like a maniac. This is moral hazard.
One line to tell them apart: before the deal, what is hidden is “who I am.” After the deal, what is hidden is “what I am doing.”
3. When buyers cannot tell good from bad, the good ones leave the market first
George Akerlof used the used car market to build a classic explanation, the “lemon market” (in American slang, a lemon is a defective car).
The logic is simple. Buyers cannot tell whether a car is any good, so they will only bid the average condition price. Sellers of good cars decide the price is not worth it and leave. The cars left in the market are worse on average, so buyers cut the price again. The better cars keep leaving. After a few rounds, only the defective cars remain, and the market may collapse entirely.
💡 This is the mechanism behind “the bad drives out the good”: it is not that bad products are more competitive. It is that good products cannot be identified, so they can only be sold at the average price, and they leave first. Hiring markets, content platforms and cross-border e-commerce supply chains keep replaying the same play.
2. Three Nobel laureates, one complete loop
The 2001 Nobel Prize in Economics went to Akerlof, Spence and Stiglitz for their work on markets with information asymmetry. Their three lines of work fit together into one loop: crisis → breaking it open → checks and balances.
| Role | Core question | Who acts |
|---|---|---|
| Akerlof: adverse selection | How does information asymmetry break a market | Nobody. The market rots on its own |
| Spence: signaling | How does the informed party prove itself | The informed side proves it voluntarily |
| Stiglitz: screening | How does the uninformed side counter | The uninformed side sets the gate |
1. Spence: signaling, where the informed side pays the cost
Michael Spence’s idea is this: if buyers cannot tell, then let the seller deliberately do something a counterfeit cannot afford to do. That something is a signal.
What makes a signal work is not that it sounds credible. It is the cost structure: the cost for a low-quality party to imitate the signal must be higher than what cheating would earn them. In the literature this is the monotone sorting condition, also called the Spence-Mirrlees condition: for the same signal, the high type pays a lower marginal cost.
A degree from an elite school works as a signal not only because of what was learned, but because it requires years of hard selection, and someone who is not capable cannot hold it up without paying too much. An expensive brand store, a no-questions-asked return promise and a performance deposit all work the same way.
2. Stiglitz: screening, where the uninformed side sets the gate
Joseph Stiglitz attacks the problem from the other end. The side with less information designs a menu of contracts, so the other side reveals its true type by choosing. This is self-selection.
An insurer offers “high deductible, low premium” and “zero deductible, high premium”. Healthy people and high-risk people sort themselves into the right plan on their own. Tiered SaaS subscriptions, stepped commission on platforms, and earn-out clauses in investment agreements are the same idea: I do not check you. I let you choose, and the choice itself is the answer.
3. How money grows out of an information gap
A large share of the excess profit in business, the economic rent, is really information rent. Not because you worked harder, but because you know one thing your opponent does not.
| Setting | The advantaged side | The disadvantaged side | How the arbitrage works |
|---|---|---|---|
| Capital markets | Core insiders, quant firms | Retail investors | First-mover information advantage and order-flow data enable low-risk arbitrage |
| Platform economy | The platform | Merchants, consumers | Exclusive black-box algorithms and full-funnel conversion data, charging a “rent tax” through auction ranking |
| Employment | Frontline staff | Management | Execution details are invisible, which can be traded for leisure or for building something hard to replace |
| Professional services | Doctors, lawyers, consultants | Clients | These are “credence goods.” After buying, the client still cannot independently judge whether they were necessary |
The conclusion is clear. Commercial competition is either building your own information barrier and collecting rent on it, or becoming the connector that lowers information asymmetry for other people. Rating agencies, search engines, audit firms and price comparison tools are all in the second group.
4. A four-step information audit
“What does each side know, and who can verify more easily?” is already a strong question. But when you drop it onto a specific deal, it splits into four steps. Pair each step with a story and it sticks better than a definition does.
1. Step one: inventory the private information. Who is hiding cards, and who is hiding actions
First separate hidden attributes from hidden actions.
Story: Lao Li’s used truck
Lao Li wants to sell his five-year-old truck to Xiao Zhang.
Lao Li knows exactly what is wrong with it. The engine has been rebuilt twice, and the transmission jams when it gets hot. What Xiao Zhang sees is only the shiny paint. This is hiding attributes before the deal.
Now suppose Xiao Zhang changes his mind and says, “Let me rent it for a year first, then decide.” Now the nervous one is Lao Li. Because the truck is in Xiao Zhang’s hands, and every day it hauls cement overloaded and takes speed bumps too fast, Lao Li cannot see that either. This is hiding actions after the deal.
Same truck, different deal structure, and the side with less information has switched. So the first step is not to ask “who is worse.” It is to ask: in this deal, what I cannot see is the other side’s cards, or the other side’s hands?
2. Step two: price the verification cost. How much does the truth cost to establish
It is cheap to suspect someone is exaggerating and expensive to get hard evidence. If verification needs specialist knowledge, a large outlay or a lot of time, you usually have to eat it.
Story: the repair shop in the middle of nowhere
You are driving across the outback and the car breaks down. The engine warning light comes on. For tens of miles there is exactly one repair shop. The owner opens the bonnet, takes one look and says, “The throttle body is gone and it burned the ECU. 3000 to fix.”
You know nothing about cars. Towing it to a 4S shop in town costs over a thousand plus a full day. Verification costs more than it is worth, so the pricing power is entirely his.
Whoever has the lower verification cost gets the louder voice at the negotiating table. This also explains a lot of pricing that “looks unfair.” The other side is not stronger. You just cannot afford to check.
3. Step three: look for signals that cannot be forged. Do not listen to what he says, look at what he stakes
A scammer saying “I am absolutely reliable” costs nothing but breath. That is cheap talk. Someone who is actually reliable will do something that would put a fake one out of business.
Story: the five-star restaurant and the roaming barbecue stall
Both owners swear their ingredients are the freshest.
The roaming vendor’s oath is worth nothing. He can pedal away at any moment and breaking his word costs almost nothing.
The restaurant downtown spent twenty million on the fit-out. What it actually communicates is: “If I serve spoiled food, sicken a customer, and get shut down by the inspectors, those twenty million are gone. Only someone planning to stay in business for years would stake that much.”
⚠️ The most common mistake is treating a claim as a signal. “We are the most professional,” “100% handmade,” “100% original.” Anyone can say those, so they carry no information at all. There is only one test: if he were cheating me, what would he lose? Treat any promise that costs him nothing as zero information.
4. Step four: design the mechanism so the other side sorts itself
If you cannot see through it and cannot be bothered to check, take another route: design a set of rules that makes anyone with something to hide opt out on their own.
Story: two renovation foremen competing on price
Both foremen brag that their work is good and that they never run late. You do not argue. You just give them two offers:
Option A: 100,000 total. Pay 80% before work starts. No penalty for delays.
Option B: 120,000 total. Pay only 30% before work starts. Pay by milestone after inspection, minus 1000 for each day of delay, and release the final payment only when everything passes.
Anyone planning to skim will invent excuses to refuse B. The genuinely skilled, efficient ones will grab B, because it pays them 20,000 more. You do not have to play detective. One set of rules makes them show themselves.
Installment payments, milestone payments, performance deposits, third-party escrow and earn-out clauses all do the same job: shift the verification cost onto the party best able to bear it.
✅ Before your next negotiation or investment, ask these four questions in order:
Is the other side hiding attributes, or hiding actions?
What will it cost me in money and time to verify this myself?
Is the cost to him of behaving badly higher than what he would gain by cheating?
Can I use payment timing, tiered pricing or escrow to move the risk to a different spot?
5. A fable: the honey market in Hive Valley
1. The story
Every year at the summer solstice, Hive Valley holds a honey market. Bees from the East Cliff and bees from the West Cliff both come to sell their own flower honey. The buyers are mole caravans from beyond the valley. They pay by the jar but never open one to taste. The jars are sealed with wax, the caravan is in a hurry, and all they have to go on is the shouting and the pattern on the jar.
East Cliff bees gather moon orchid from the cliff tops. The nectar is clear as dew, and one jar takes them across thirty li of cliff face and seven days of flying. West Cliff bees gather wild chrysanthemum from the valley floor. The nectar is cloudy and cloyingly sweet, and one jar takes half a day. Both jars carry the same golden spiral pattern, the valley’s inherited mark of “premium” quality.
At first the moles offered ten pine nuts for a jar. The East Cliff bees thought that was fair, and the West Cliff bees thought it was a bargain. Then the caravan took the jars home, opened them, and found that six out of ten were wild chrysanthemum honey. The following year they offered only five pine nuts, pricing at the average quality.
The East Cliff bees were angry. Five pine nuts does not even cover the energy of flying across the cliff. So they switched to making wild chrysanthemum honey too. Same jars, far lower cost. By the third year the market was almost entirely wild chrysanthemum, the moles were tired of telling it apart, and the price dropped to three pine nuts. The East Cliff bees left entirely, and the valley had nothing left but the cloying cloudy honey.
One old East Cliff bee would not accept it, and went to find the wise owl. The owl said, “You have to make buyers believe you really flew the cliff. But saying it does nothing, because the West Cliff bees will say it too.”
The old bee thought all night. In the morning he raised a tall spruce pole in the market square with a moon orchid hanging from the top, and announced: “Any buyer who buys a jar from me gets carried up to the top of this pole, and can see the flower fields on the cliff with his own eyes. Only I am strong enough to carry him. I have trained my wings for thirty years. The West Cliff bees have short wings. They cannot carry a single pine nut.”
One young mole paid ten pine nuts. The old bee carried him up the pole, shaking all the way, and from the top the mole saw the silver-glinting sea of flowers on the distant cliffs. He came down telling everyone. From then on the old bee carved a “witnessed in person” mark onto a stick for every buyer, and the price of honey went back to ten pine nuts, then above it.
The West Cliff bees raised a pole too, but they started shaking after half a pine nut and fell three times. The buyers laughed: “If you cannot even carry half a pine nut, how could you gather flowers on a cliff?”
The old bee’s business was not easy either. Every jar cost him a full carry, and he could only sell three jars a day. A young bee asked him, “Why not simplify? Why raise the pole but not carry people?” The old bee shook his head. “If it does not take real strength, the West Cliff bees can put on the act too, and then this pole means nothing.”
Years later the market split into tiers again. High-pole honey and lowland honey each sold at their own price. And the moles had learned one lesson: when you cannot tell where honey comes from, the good honey disappears. The only way to bring it back is to make the seller pay a cost that a fake seller cannot pay.
2. What this story maps onto
| Story element | Concept it maps to |
|---|---|
| East Cliff bees / West Cliff bees | High-quality and low-quality sellers |
| The mole caravan never opens jars | Buyers cannot verify quality in advance |
| The golden spiral pattern | Cheap signal: anyone can copy it, so it carries no information |
| Ten → five → three pine nuts | Lemon market dynamics produced by average pricing |
| Raising the pole and carrying buyers up | A signal that costs something |
| West Cliff bees cannot carry half a pine nut | Mimicking is so expensive for the low type that they give up |
| The “witnessed in person” stick | The verifiable result of a signal, which is not itself the signal |
| The proposal to skip carrying people | Lowering the signal’s cost, which breaks the separation |
The last row is worth pulling out on its own. As soon as a signal can be copied cheaply, it collapses back into a golden spiral. So a working signal has to satisfy two conditions: the high-quality party’s net gain from sending it is higher than from not sending it, and the low-quality party’s net gain from imitating it is lower than from not imitating. The old bee’s wings are the result of thirty years. The West Cliff bees were born with short wings and cannot fake that in the short run. That cost gap makes the two kinds of bees choose different actions, which lets the buyer infer the type, which is what economics calls a separating equilibrium.
Education works as a signal in the labor market, and dividends work as a signal in corporate finance, for exactly the same reason.
6. Using the four steps in two real situations
1. I did a lot of work and my boss does not know
This is a textbook principal-agent problem, with a layer of black-box execution on top.
1) What private information are you holding
There is a lot of invisible work in operations: keyword filtering, broken-link fixes, competitor monitoring, asset tuning. You know your own load. The boss sees only the final numbers — inquiry volume, traffic curve. He cannot tell the difference between “I spent four hours polishing ten core pages” and “I slacked off for four hours and clicked a couple of things.” When the process is invisible, people’s default assumption leans toward incompetence.
2) Why the boss does not want to understand how hard operations are
Because his verification cost is too high. His attention is on cash flow, supply chain and business development. Making him learn SEO tag rules, plugin code or ad bidding algorithms costs far more than it returns. So he retreats to the cheapest available way to evaluate: look only at visible output, or guess outright — “how long does it take to publish a few articles and change a few words?“
3) What signals cannot be forged
Saying “I was exhausted today” in the group chat is cheap talk, and can even be read as “they are only slow because they are bad.” What you need instead is a deliverable that is hard to fake:
- Structured numbers: turn “it took a long time” into “checked 320 broken links, cleaned up 1,200 invalid keywords, restructured page metadata on 15 pages.”
- A comparison baseline: give a reference frame, for example “sites of this size normally need three weeks for this kind of recovery. We used a script to get it down to four days.”
4) How to use the mechanism to turn the black box white
Do not wait for the boss to come and understand you. Make the workload surface on its own:
1)Confirm the schedule up front: when a request looks simple, do not just reply “noted.” Reply with “I’ll split it into A, B and C. A is testing the API, B is cleaning the data, estimated 6 hours, delivered Wednesday afternoon.” This step directly lowers his cost of comprehension.
2)A process board: turn the underwater actions into cards that move on a board, so blockers and throughput are always visible.
3)Report return after the fact: do not report how hard it was. Report “because I did these four things, I avoided this risk and improved this much efficiency.”
Your boss not knowing the details is an unavoidable consequence of division of labor. Converting invisible labor into a visible asset is your own job.
2. My product really is better, and my article really was not written by AI
The marginal cost of AI-generated content is heading toward zero, so the internet is filling up with homogenized content that has tidy formatting and safely moderate opinions. That is a new lemon market.
1) What each side knows
You know you did original research, tested repeatedly, and folded in years of lessons learned the hard way. It took two full days. The reader does not know that, and he has been fooled by “batch content laundering” too many times, so he has a defensive bias: he sees a tidy long article and assumes it is filler before reading a single line.
2) Why readers cannot be bothered to tell
Reading time is the scarce resource. He cannot read the whole thing in three seconds to judge its quality, and AI can just as easily generate subheadings, bold text and structure. Judging word by word whether the argument is original costs an enormous amount of verification effort. Since verifying is tiring, he draws one line through it: treat all of it as average-quality and scroll past.
3) What AI cannot forge
Writing “this article was written entirely by a human” at the end is cheap talk. Content farm accounts write it too. The real signals are:
- First-hand process that leaves traces: screenshots you took yourself, real error logs, communication records with timestamps. A general model can invent a plausible explanation. It cannot invent the specific parameter and the odd value you ran into yesterday.
- Original judgments specific enough to sting: AI’s core function is maximizing probability across the whole web, so its opinions tend to be bland and grand. Offering a non-consensus call, backed by obscure but lethal field detail, is very hard to imitate.
- Real reputation as collateral: a real name, a traceable track record, a long-term personal brand. If the content is fake, that reputation depreciates permanently.
4) How to design the mechanism
Do not make the reader reach the end of the article to discover there was anything real in it. Move the verification gate to the front:
1)Put unfakeable evidence in the first screen — an original test screenshot, or a conclusion that goes against common sense.
2)Write less of “in principle you should pay attention to…” and more of “last Tuesday when I adjusted parameter X I hit error Y.” A concrete first-person scene is itself a signal.
3)Attach something that can be verified immediately: an in-house template, a retrospective spreadsheet, an original code snippet.
7. Three things worth taking away
- Information asymmetry is the default state of a transaction. First work out whether the other side is hiding an attribute or an action.
- Pricing power follows verification cost. What you cannot afford to check, you have to buy at their price.
- The value of a promise is not in its wording, it is in what is staked behind it. Only a cost a counterfeit cannot pay counts as evidence.
So the next time you analyze a deal, do not stop at “who knows what.” Take one step further and ask: who can send a signal their opponent cannot imitate, at a low enough cost? That is usually where the profit sits.